{
  "markdown": "# Digital Innovation Agents\n\n> **When code costs almost nothing, the plan becomes the product.**\n>\n> A V-Model workflow that walks your AI coding assistant through Business\n> Analysis, Requirements Engineering, Architecture, Coding, Testing, and\n> a Security Audit, with quality-gated handoffs between every phase.\n\n**Full documentation:** [pssah4.github.io/digital-innovation-agents](https://pssah4.github.io/digital-innovation-agents/)\n\nShipping code is a solved problem. What most teams still lack is evidence\nthat the features they ship matter to a real user. Digital Innovation\nAgents pair a battle-tested innovation methodology with a state-of-the-art\ncoding workflow, so your AI never builds the wrong thing at speed. Works\nacross **Claude Code**, **Cursor**, **Codex**, **OpenCode**, **Gemini CLI**,\nand **GitHub Copilot**.\n\n<p align=\"center\">\n  <img src=\"https://raw.githubusercontent.com/pssah4/digital-innovation-agents/main/docs/public/v-model-overview.svg\" alt=\"V-Model workflow for AI coding assistants: six phases (Business analysis, Requirements engineering, Architecture, Coding, Testing, Security audit) plus a Closing handoff. Two consistency buses run beneath the phases: BACKLOG.md as status source of truth and ARCHITECTURE.map as code source of truth. Four dashed feedback loops show test fix, mid-course discovery, security fix, and living-documents writeback.\" width=\"100%\" />\n</p>\n\n---\n\n## What this is\n\nThe project ships thirteen skills that run inside your AI coding\nassistant. Six are V-Model phase skills (business analysis,\nrequirements engineering, architecture, coding, testing, security\naudit). One is the entry point for non-greenfield projects\n(`/dia-realign`). One is the on-demand workflow guide (`/dia-guide`,\nan explicit command the model never auto-invokes). Five are\nfoundation skills (project conventions, consistency check, humanizer,\ndia-setup, dia-bootstrap). Every phase skill owns one part of the\nV-Model, has its own quality gates, and hands off a structured\nartifact to the next phase; the transition record lives as DIA\ntrailers (`DIA-Phase`, `DIA-Handoff`, `DIA-Triage`) on the phase-end\ncommits, readable with plain git. Every decision stays traceable\nfrom a real user problem through requirements, architecture, code,\ntests, and a security audit.\n\nTwo entry points cover greenfield and everything else:\n\n- **Greenfield:** `/business-analysis` starts with structured discovery\n  (users, needs, insights, critical hypotheses) and walks forward\n  through the V-Model.\n- **Brownfield and legacy DIA:** `/dia-realign` detects the repo state\n  and picks the fitting mode. For a codebase without artifacts it\n  walks the V backwards and produces a wayfinder, post-hoc ADRs, an\n  arc42 reference, a FEATURE inventory, a backlog seed, and an\n  evidence-based BA draft (every claim sourced to a file path or doc\n  section, nothing invented). For an older DIA project it runs the\n  idempotent migration script pass (status drift, ID schemas,\n  analysis/ flattening, backlog regeneration) and then fills the\n  remaining gaps.\n\n## Innovation methodology, not just automation\n\nThe BA and RE agents ship a catalog of 32 innovation methods (qualitative\ninterviews, extreme users, fly on the wall, cultural probes, persona\nsynthesis, stakeholder maps, jobs to be done, brainwriting, TRIZ, wizard\nof oz, pre-mortem, value proposition quantification, and more) organised\nas [method cards in the docs](https://pssah4.github.io/digital-innovation-agents/reference/methods-discovery).\n\nDuring a BA or RE session, when your answers go thin, the agent stops\nthe interview and proposes the matching field method with a one-page\ncard: what it produces, when to reach for it, how to run it, team and\ntime, things that go wrong, and what to bring back to the session. The\nactual research work stays human-to-human: interviews with real users,\nobservations in the real context, prototypes on real hands. The agent's\njob is to spot the gap and pick the right method, not to replace the\nwork.\n\n## Quick start\n\nPick your platform. Each installer drops the same thirteen skills with\nthe same templates and quality gates into your tool of choice.\n\n### Claude Code (recommended)\n\n`/plugin` lives in the **Claude Code CLI**, not in the VS Code or\nJetBrains extensions. If `claude --version` returns\n`command not found`, install the CLI first:\n\n```bash\ncurl -fsSL https://claude.ai/install.sh | bash    # official installer\n# or: brew install --cask claude-code              # macOS Homebrew\n# or: npm install -g @anthropic-ai/claude-code     # any OS with Node\n```\n\nReopen your shell (`source ~/.zshrc` or `~/.bashrc`), then install the\nplugin:\n\n```bash\nclaude\n```\n\n```\n/plugin marketplace add https://github.com/pssah4/digital-innovation-agents.git\n/plugin install digital-innovation-agents@pssah4-skills\n```\n\nType `/` in any new session to see the skills in autocomplete. The\n`dia-bootstrap` skill loads automatically at session\nstart as a brief orientation.\n\n**VS Code, JetBrains, and Cursor extensions cannot install plugins.**\nRunning `/plugin marketplace add ...` inside the VS Code Claude Code\nextension returns `/plugin isn't available in this environment`.\nInstall once through the CLI as above. The skills land under\n`~/.claude/skills/` and the IDE extension picks them up from the same\nglobal directory on the next session start. On Windows without WSL,\nthe CLI is experimental; install through WSL or copy the manually:\n\nThe manual install copies the **complete plugin bundle** (skills,\ntools, hooks, scripts) under a stable path and symlinks the skills\ninto `~/.claude/skills/`. Skills invoke tooling at\n`${DIA_PLUGIN_ROOT}/tools/...`, so `DIA_PLUGIN_ROOT` is exported\nfrom the install location. Re-run the block to update; it pulls\nthe latest commit and rewrites the bundle in place. Skills\nrenamed or removed in newer DIA versions (for example\n`dia-orchestrator` from v2) are deleted explicitly so no stale\nskill folders survive the upgrade.\n\n```bash\n# Stable plugin location. Override with DIA_PLUGIN_ROOT env if needed.\nDIA_PLUGIN_ROOT=\"${DIA_PLUGIN_ROOT:-$HOME/.local/share/dia-plugin}\"\n\n# Clone or update the plugin bundle\nif [ -d \"$DIA_PLUGIN_ROOT/.git\" ]; then\n  git -C \"$DIA_PLUGIN_ROOT\" fetch --tags --prune\n  git -C \"$DIA_PLUGIN_ROOT\" reset --hard origin/main\nelse\n  mkdir -p \"$(dirname \"$DIA_PLUGIN_ROOT\")\"\n  rm -rf \"$DIA_PLUGIN_ROOT\"\n  git clone https://github.com/pssah4/digital-innovation-agents.git \"$DIA_PLUGIN_ROOT\"\nfi\n\nmkdir -p ~/.claude/skills\n\n# Remove legacy DIA skills that were renamed or dropped\nfor legacy in dia-orchestrator reverse-engineering dia-migration; do\n  rm -rf \"$HOME/.claude/skills/$legacy\"\ndone\n\n# Symlink the current DIA skill set (covers future renames automatically)\nfor skill in project-conventions dia-realign business-analysis \\\n             requirements-engineering architecture coding testing \\\n             security-audit consistency-check humanizer dia-guide \\\n             dia-setup dia-bootstrap; do\n  rm -rf \"$HOME/.claude/skills/$skill\"\n  ln -sfn \"$DIA_PLUGIN_ROOT/skills/$skill\" \"$HOME/.claude/skills/$skill\"\ndone\n\n# Persist DIA_PLUGIN_ROOT so skills can resolve tools/ at runtime\nshell_rc=\"$HOME/.zshrc\"\n[ -f \"$HOME/.bashrc\" ] && shell_rc=\"$HOME/.bashrc\"\nif ! grep -q \"DIA_PLUGIN_ROOT=\" \"$shell_rc\" 2>/dev/null; then\n  echo \"export DIA_PLUGIN_ROOT=\\\"$DIA_PLUGIN_ROOT\\\"\" >> \"$shell_rc\"\nfi\nexport DIA_PLUGIN_ROOT\n```\n\nAfter the first install, open a new shell so `DIA_PLUGIN_ROOT` is\nset, then start `claude`. Skills resolve their helper scripts at\n`$DIA_PLUGIN_ROOT/tools/...` regardless of the user-project cwd.\n\n### Cursor\n\n```\n/add-plugin digital-innovation-agents\n```\n\nOr search for \"digital-innovation-agents\" in the Cursor plugin\nmarketplace.\n\n### GitHub Copilot (CLI and VS Code)\n\nGitHub Copilot has no marketplace command. Install by copying the\n`.github/` directory plus the helper tools into your project. The\nagents call `flow.py`, `anchor.py`, the migration scripts, and the\nconsistency check, so `tools/` and `scripts/` must be available\nlocally.\n\nRe-run the block to update; the source checkout is pulled to the\nlatest commit and each target subfolder is wiped before copy, so\nno stale Copilot agents, chat modes, or helper scripts survive an\nupgrade.\n\n```bash\n# Clone or update the source checkout\nif [ -d /tmp/dia/.git ]; then\n  git -C /tmp/dia fetch --tags --prune\n  git -C /tmp/dia reset --hard origin/main\nelse\n  rm -rf /tmp/dia\n  git clone https://github.com/pssah4/digital-innovation-agents.git /tmp/dia\nfi\n\nmkdir -p .github\n\n# Wipe old DIA copies before installing the current set\nfor sub in agents chatmodes instructions templates; do\n  rm -rf \".github/$sub\"\n  cp -r \"/tmp/dia/.github/$sub\" \".github/$sub\"\ndone\ncp /tmp/dia/.github/copilot-instructions.md .github/copilot-instructions.md\n\n# Install the helper tools (flow.py, anchor.py, migration, hooks)\n# at the project root so the agents can invoke them\nfor sub in tools scripts hooks; do\n  rm -rf \"$sub\"\n  cp -r \"/tmp/dia/$sub\" \"$sub\"\ndone\n\n# Skills resolve tools/ relative to the project root in this layout,\n# so DIA_PLUGIN_ROOT points at the project itself\necho 'export DIA_PLUGIN_ROOT=\"$(pwd)\"' >> .envrc 2>/dev/null || true\n```\n\nThe Copilot install brings the helper tools into the project rather\nthan to a global location because Copilot agents run with the\nproject as their working directory and have no plugin-bundle path\nto fall back on.\n\nCopilot Chat picks the agents up automatically on the next session. In\nCopilot Chat:\n\n```\n@business-analyst I want to build a tool that helps teams run better retrospectives\n@requirements-engineer Here is my BA document, create epics and features\n@architect Design the architecture based on the requirements handoff\n@developer Implement the first feature\n@debugger Tests are failing, analyze the error log\n```\n\nThe Copilot agents run the same Exploration / Ideation / Validation\ncycle, the same templates, and the same quality gates as the Claude\nCode skills.\n\n### Codex\n\nTell Codex:\n\n```\nFetch and follow instructions from https://raw.githubusercontent.com/pssah4/digital-innovation-agents/main/.codex/INSTALL.md\n```\n\nDetailed docs: [.codex/INSTALL.md](.codex/INSTALL.md)\n\n### OpenCode\n\nTell OpenCode:\n\n```\nFetch and follow instructions from https://raw.githubusercontent.com/pssah4/digital-innovation-agents/main/.opencode/INSTALL.md\n```\n\nDetailed docs: [.opencode/INSTALL.md](.opencode/INSTALL.md)\n\n### Gemini CLI\n\n```bash\ngemini extensions install https://github.com/pssah4/digital-innovation-agents\n```\n\nTo update:\n\n```bash\ngemini extensions update digital-innovation-agents\n```\n\n## Verify the install\n\nStart a session in your chosen platform and try one of these:\n\n```\n/dia-setup                 Activate the workflow in this project\n/dia-guide                 Orientation read: state audit and next-phase recommendation\n/business-analysis         Start a structured business analysis\n/dia-realign               Brownfield entry and legacy DIA upgrade\n```\n\n`/dia-setup` is the first call in any new project. It asks for the\nmode (`off`, `git-only`, or `github-sync`) and the profile (`full`\nor `lean`), writes `.dia/config.toml`, and adds a managed anchor\nblock to your existing `CLAUDE.md`, `AGENTS.md`, `GEMINI.md`,\n`.cursorrules`, or similar agent files. Re-run any time to change\nthe mode or profile, or to remove the anchor.\n\nOr ask a natural-language question like \"help me analyse this business\nproblem\". The agent should invoke the matching skill.\n\nTroubleshooting:\n\n- **Claude Code or Cursor:** restart the session, the SessionStart hook\n  loads the skill overview automatically.\n- **Codex:** verify the symlink with `ls -la ~/.agents/skills/digital-innovation-agents`.\n- **OpenCode:** check logs with `opencode run --print-logs \"hello\" 2>&1 | grep -i digital-innovation`.\n- **Gemini CLI:** run `gemini extensions list`.\n\n## The skills\n\nThe thirteen skills split into three groups: V-Model phase skills\n(the ones that own a phase or move you between phases), foundation\nskills (rules and consistency), and the bootstrap skill\n(`dia-bootstrap` loads on session start to introduce the workflow).\n\n### V-Model phase skills\n\n| Phase | What it does | Claude Code | Copilot |\n|---|---|---|---|\n| **DIA Realign** | One entry point for brownfield codebases and legacy DIA repos. Detects the repo state, then runs a full reverse walk, the migration script pass, or a gap walk. Every claim sourced, idempotent, branch-safe. | `/dia-realign` | `@reverse-engineer` |\n| **Business Analysis** | Exploration, Ideation, and Validation cycle with structured interviews, probing techniques, and the 32-method discovery catalog. Condenses the dialog into a 40-line BA record. | `/business-analysis` | `@business-analyst` |\n| **Requirements Engineering** | Epics, FEAT-EE-FF features, tech-agnostic success criteria, user stories across functional / emotional / social levels, critical hypotheses. | `/requirements-engineering` | `@requirements-engineer` |\n| **Architecture** | ADRs with kinds (post-hoc as the normal case, choice, constraint) and the abstraction rule (no code paths in core sections), arc42 constraints doc, navigation artifacts (SYSTEM-MAP, decisions router), wayfinder maintenance, plan-context ref index. | `/architecture` | `@architect` |\n| **Coding** | Context handoff, critical review against the real codebase, PLAN-NN persistence with coverage gate, TDD by default (opt-out `--no-tdd` with user confirmation), bug-capture entry, artifact writeback during implementation. | `/coding` | `@developer` |\n| **Testing** | Unit and integration tests with the AAA pattern, FIRST principles, coverage targets, and a fix-loop until green. Test edits under \"the spec changed\" require three pieces of evidence. | `/testing` | built-in |\n| **Security Audit** | OWASP Top 10, LLM Top 10, SAST, SCA, supply-chain checks, Zero Trust review with a fix-loop. Two modes: per-item audit and periodic full-codebase audit. | `/security-audit` | `@security-auditor` |\n| **V-Model Workflow Guide** | Explicit orientation command (never auto-invoked): reads project state and the DIA commit trailers, recommends the next phase skill, audits the last phase-end commit, and emits the Closing Handoff after a green security audit. The guide does not drive transitions; phase skills are autonomous. | `/dia-guide` | built-in |\n| **Debugging** | Root-cause analysis, systematic error resolution, causal chain documentation. Bugs land as FIX-EE-FF-NN rows in the backlog plus detail files in `_devprocess/requirements/fixes/`. | default agent | `@debugger` |\n\n### Foundation skills\n\n| Skill | What it does | Claude Code |\n|---|---|---|\n| **Project Conventions** | Three-layer documentation model (Wayfinder, Rule sets, Backlog, Detail artifacts), directory structure, naming standards, writing-style rules. | `/project-conventions` |\n| **Consistency Check** | Explicit command that verifies the V-Model artifact graph: dead links, orphan features, status drift, missing references. Modes A (syntactic), B (semantic), C (interactive fix-loop). Mandatory once per cycle before release; the pre-commit hook covers the drift-critical invariants between runs. | `/consistency-check` |\n| **Humanizer** | Strips AI vocabulary, em dashes, negative parallelisms, and filler from every artifact. Enforces sentence case and active voice. | `/humanizer` |\n| **DIA Bootstrap** | Loads automatically on session start. Carries the entry-point catalog, helper-script path resolution rule, activation contract, opt-out behaviour. Not invoked manually. | `dia-bootstrap` |\n\n## Scope levels\n\nThe skills adapt their depth to your project scope. Match the tier to\nthe size of the question.\n\n| Scope | Exploration | Ideation | Validation | Typical duration |\n|---|---|---|---|---|\n| **Simple Test** | Minimal (user and problem) | Describe the solution | Skip | Hours to 1-2 days |\n| **Proof of Concept** | Shortened (user, needs, HMW) | Full | Hypotheses and feasibility | 1-4 weeks |\n| **MVP** | Full 10-section Exploration board | Full | Full market assessment | 2-6 months |\n\nA Simple Test does not need a stakeholder map. An MVP does not get away\nwithout one.\n\n## The lean profile\n\nBesides the workflow mode, `.dia/config.toml` carries a `profile`\nfield: `full` (default) or `lean`. The lean profile makes only three\nthings binding: durable decisions, stable navigation, and backlog\nstatus. Rules live consolidated in AGENTS.md (CLAUDE.md points at\nit), navigation lives in `_devprocess/SYSTEM-MAP.md`, decisions are\npost-hoc ADRs behind a `decisions/README.md` router table, and\nstatus lives in GitHub Issues (github-sync) or a thin BACKLOG\n(git-only). All other phase skills stay available but advisory.\nPick lean when the team will not run BA/RE ceremony anyway; a thin\nlayer that is maintained beats a full layer that drifts.\n\n## Tech-agnostic requirements\n\nSuccess Criteria stay free of technology vocabulary. No OAuth, REST,\nPostgreSQL, or React in the contract between the user and the team.\nTechnical details live in a separate Technical NFRs section and in the\nADRs that follow in `/architecture`. See the\n[Tech-agnostic Requirements](https://pssah4.github.io/digital-innovation-agents/concepts/tech-agnostic-requirements)\npage in the docs for the full ruleset.\n\n## Living documents\n\nADRs, features, architecture docs, and the backlog update continuously\nduring implementation. At release time, documentation reflects what was\nactually built, not what was originally planned. The\n`_devprocess/context/BACKLOG.md` file is the single source of truth\nfor project state, and every phase skill touches it in the same edit\npass as the code it affects.\n\n## Design principles\n\n1. Understand the problem before designing the solution.\n2. Separate what the system does (user-observable, tech-free) from how\n   it does it (ADRs, NFRs).\n3. Propose the right research method instead of grinding through a\n   question list when your answers go thin.\n4. No phase proceeds until its quality gate is met.\n5. Every agent reads the real codebase before producing output. The\n   project's `CLAUDE.md` always takes precedence over generic skill\n   instructions.\n\n## Documentation\n\nEvery guide, tutorial, concept page, and method card lives at\n[pssah4.github.io/digital-innovation-agents](https://pssah4.github.io/digital-innovation-agents/).\n\nStart here:\n\n- [Your first Business Analysis tutorial](https://pssah4.github.io/digital-innovation-agents/tutorials/first-business-analysis)\n- [A full V-Model run](https://pssah4.github.io/digital-innovation-agents/tutorials/full-v-model-run)\n- [Discovery methods](https://pssah4.github.io/digital-innovation-agents/reference/methods-discovery),\n  [Ideation methods](https://pssah4.github.io/digital-innovation-agents/reference/methods-ideation),\n  [Validation methods](https://pssah4.github.io/digital-innovation-agents/reference/methods-validation)\n- [DIA Realign guide](https://pssah4.github.io/digital-innovation-agents/guides/dia-realign)\n\n## Versions\n\n| Version | Status | Install |\n|---|---|---|\n| **v3** (main) | Active, recommended. Three-layer documentation model, FEAT-EE-FF IDs, FIX/IMP detail files, PLAN-NN persistence, GitHub flow.py integration, subtype-aware Done-definition. | See Quick start above |\n| **v2.x** | Frozen snapshot, no longer maintained | `git clone --branch v2.4.0 https://github.com/pssah4/digital-innovation-agents.git` |\n| **v1.0.0** | Frozen snapshot, no longer maintained | `git clone --branch v1.0.0 https://github.com/pssah4/digital-innovation-agents.git` |\n\nSee [CHANGELOG.md](CHANGELOG.md) for details. Existing v1 or v2 projects\nupgrade through `/dia-realign` (Mode B). v1 and v2 are historical\nsnapshots and not actively maintained; for current behaviour use the\nmarketplace or platform-specific install on v3.\n\n## License\n\nMIT License. Copyright (c) 2025 Sebastian Hanke. See [LICENSE](LICENSE).\n\n## Acknowledgments\n\nBuilt with:\n\n- Claude Code Skills, Claude Agent SDK, and GitHub Copilot Agents\n- Innovation methodology from design thinking and lean startup practice\n- Jobs-to-be-Done framework\n- arc42 architecture documentation template\n- MADR (Markdown Architectural Decision Records)\n- OWASP Top 10 and LLM Top 10\n- AAA pattern and FIRST principles for testing\n",
  "bytes": 20225,
  "sha": "701b127f375f7d0b9f0a8a569cdeab1d2c091c0e6dd8a07123a177ebe54767ec",
  "repo_slug": "pssah4/digital-innovation-agents",
  "fonte": "repo",
  "truncated": false,
  "api": "https://agentalog.com/api/listings/plg_pssah4_digital_innovation_agents_ee1452af/readme"
}